(CRITICS): an International, Open-Label, Randomised Phase 3 Trial

Articles

Chemotherapy versus chemoradiotherapy after surgery and preoperative chemotherapy for resectable gastric cancer (CRITICS): an international, open-label, randomised phase 3 trial

Annemieke Cats*, Edwin P M Jansen*, Nicole C T van Grieken, Karolina Sikorska, Pehr Lind, Marianne Nordsmark, Elma Meershoek-Klein Kranenbarg, Henk Boot, Anouk K Trip, H A Maurits Swellengrebel, Hanneke W M van Laarhoven, Hein Putter, Johanna W van Sandick, Mark I van Berge Henegouwen, Henk H Hartgrink, Harm van Tinteren, Cornelis J H van de Velde†, Marcel Verheij†, for the CRITICS investigators‡

Summary Lancet Oncol 2018; 19: 616–28 Background Both perioperative chemotherapy and postoperative chemoradiotherapy improve survival in patients Published Online with resectable gastric cancer from Europe and North America. To our knowledge, these treatment strategies have not April 9, 2018 been investigated in a head to head comparison. We aimed to compare perioperative chemotherapy with preoperative http://dx.doi.org/10.1016/ chemotherapy and postoperative chemoradiotherapy in patients with resectable gastric adenocarcinoma. S1470-2045(18)30132-3 See Comment page 581 Methods In this investigator-initiated, open-label, randomised phase 3 trial, we enrolled patients aged 18 years or older *Contributed equally who had stage IB– IVA resectable gastric or gastro-oesophageal adenocarcinoma (as defined by the American Joint †Contributed equally Committee on Cancer, sixth edition), with a WHO performance status of 0 or 1, and adequate cardiac, bone marrow, ‡A complete list of CRITICS liver, and kidney function. Patients were enrolled from 56 hospitals in the Netherlands, Sweden, and Denmark, and investigators is provided in the appendix were randomly assigned (1:1) with a computerised minimisation programme with a random element to either Department of Gastrointestinal perioperative chemotherapy (chemotherapy group) or preoperative chemotherapy with postoperative chemoradiotherapy Oncology (A Cats MD, (chemoradiotherapy group). Randomisation was done before patients were given any preoperative chemotherapy H Boot MD, treatment and was stratified by histological subtype, tumour localisation, and hospital. Patients and investigators were H A M Swellengrebel MD), not masked to treatment allocation. Surgery consisted of a radical resection of the primary tumour and at least a D1+ Department of Radiation Oncology (E P M Jansen MD, lymph node dissection. Postoperative treatment started within 4–12 weeks after surgery. Chemotherapy consisted of A K Trip MD, Prof M Verheij MD), three preoperative 21-day cycles and three postoperative cycles of intravenous epirubicin (50 mg/m² on day 1), cisplatin Department of Biometrics (60 mg/m² on day 1) or oxaliplatin (130 mg/m² on day 1), and capecitabine (1000 mg/m² orally as tablets twice daily for (K Sikorska PhD, 14 days in combination with epirubicin and cisplatin, or 625 mg/m² orally as tablets twice daily for 21 days in H van Tinteren PhD), and Department of Surgery combination with epirubicin and oxaliplatin), received once every three weeks. Chemoradiotherapy consisted of 45 Gy (J W van Sandick MD), in 25 fractions of 1·8 Gy, for 5 weeks, five daily fractions per week, combined with capecitabine (575 mg/m² orally Netherlands Cancer Institute, twice daily on radiotherapy days) and cisplatin (20 mg/m² intravenously on day 1 of each 5 weeks of radiation Amsterdam, Netherlands; treatment). The primary endpoint was overall survival, analysed by intention-to-treat. The CRITICS trial is registered Department of Pathology, VU Medical Center, Amsterdam, at ClinicalTrials.gov, number NCT00407186; EudraCT, number 2006-004130-32; and CKTO, 2006-02. Netherlands (N C T van Grieken MD); Findings Between Jan 11, 2007, and April 17, 2015, 788 patients were enrolled and randomly assigned to chemotherapy Department of Medical (n=393) or chemoradiotherapy (n=395). After preoperative chemotherapy, 372 (95%) of 393 patients in the chemotherapy Oncology (Prof H W M van Laarhoven MD) group and 369 (93%) of 395 patients in the chemoradiotherapy group proceeded to surgery, with a potentially curative and Department of Surgery resection done in 310 (79%) of 393 patients in the chemotherapy group and 326 (83%) of 395 in the chemoradiotherapy (Prof M I van Berge Henegouwen group. Postoperatively, 233 (59%) of 393 patients started chemotherapy and 245 (62%) of 395 started chemoradiotherapy. MD), Academic Medical Center, At a median follow-up of 61·4 months (IQR 43·3–82·8), median overall survival was 43 months (95% CI 31–57) in the Amsterdam, Netherlands; Department of Surgery chemotherapy group and 37 months (30–48) in the chemoradiotherapy group (hazard ratio from stratified analysis (E Meershoek-Klein Kranenbarg 1·01 (95% CI 0·84–1·22; p=0·90). After preoperative chemotherapy, in the total safety population of 781 patients MSc, H H Hartgrink MD, (assessed together), there were 368 (47%) grade 3 adverse events; 130 (17%) grade 4 adverse events, and 13 (2%) deaths. Prof C J H van de Velde MD) and Department of Medical Causes of death during preoperative treatment were diarrhoea (n=2), dihydropyrimidine deficiency (n=1), sudden Statistics (Prof H Putter PhD), death (n=1), cardiovascular events (n=8), and functional bowel obstruction (n=1). During postoperative treatment, Leiden University Medical grade 3 and 4 adverse events occurred in 113 (48%) and 22 (9%) of 233 patients in the chemotherapy group, respectively, Center, Leiden, Netherlands; and in 101 (41%) and ten (4%) of 245 patients in the chemoradiotherapy group, respectively. Non-febrile neutropenia Department of Oncology and Pathology, Karolinska occurred more frequently during postoperative chemotherapy (79 [34%] of 233) than during postoperative Institute, Stockholm, Sweden chemoradiotherapy (11 [4%] of 245). No deaths were observed during postoperative treatment. (P Lind MD); and Department of Oncology, Aarhus University Interpretation Postoperative chemoradiotherapy did not improve overall survival compared with postoperative Hospital, Aarhus, Denmark (M Nordsmark MD) chemotherapy in patients with resectable gastric cancer treated with adequate preoperative chemotherapy and surgery. In view of the poor postoperative patient compliance in both treatment groups, future studies should focus on optimising preoperative treatment strategies.

616 www.thelancet.com/oncology Vol 19 May 2018 Articles

Funding Dutch Cancer Society, Dutch Colorectal Cancer Group, and Hoffmann-La Roche. Correspondence to: Dr Edwin Jansen, Netherlands Copyright © 2018 Elsevier Ltd. All rights reserved. Cancer Institute, 1066 CX, Amsterdam, Netherlands [email protected] Introduction We designed the ChemoRadiotherapy after Induction See Online for appendix Gastric cancer ranks fifth in cancer mortality globally.1 In chemoTherapy In Cancer of the Stomach (CRITICS) patients from Europe and North America with localised study to incorporate both treatment strategies. The or locally advanced disease, prognosis remains dismal study addresses the question of whether postoperative after surgery alone, with most patients relapsing chemoradiotherapy improves survival as compared within 2 years after treatment. Extended lymph node with postoperative chemotherapy in patients who are dissections improve cancer-specific survival, but also treated with preoperative chemotherapy followed by increase surgical morbidity and mortality.2,3 surgery. Because oxaliplatin and capecitabine have In 2001, the US Intergroup 0116 trial4 reported that equal efficacy and better tolerability than cisplatin postoperative fluorouracil monotherapy in combination and fluorouracil in advanced gastric cancer,7 we with fluorouracil-based chemoradiotherapy improved substituted fluorouracil with capecitabine and used both locoregional control and overall survival compared either cisplatin or oxaliplatin in the combination with surgery alone in medically fit patients who had a chemotherapy regimens. microscopically radical resection.4 This beneficial effect persisted after more than 10 years of follow-up.5 Methods Meanwhile, results from the British Medical Research Study design and participants Council Adjuvant Gastric Infusional Chemotherapy The CRITICS study is an investigator-initiated, open- (MAGIC) study showed that perioperative combination label, randomised phase 3 trial, with patients recruited chemotherapy with epirubicin, cisplatin, and fluoro from the Netherlands (44 hospitals), Sweden (11 hospitals), uracil was associated with tumour downsizing and and Denmark (one hospital; appendix pp 1–2). A detailed downstaging, and a significant overall survival benefit description of the study has been published previously.8 compared with surgery alone.6 The study protocol is provided in the appendix (pp 3–65).

Research in context Evidence before this study adjuvant treatment in patients with resectable gastric cancer. In March 15, 2006, when writing the study protocol, we However, because of the previously reported tumour searched PubMed and abstracts of the main annual oncology downsizing and downstaging potential of chemotherapy, all meetings for publications relating to adjuvant and neoadjuvant patients in this study were treated with preoperative chemotherapy and chemoradiotherapy in patients with chemotherapy. Results of this trial are therefore of high clinical resectable gastric cancer. We limited this search to articles relevance for current practice in Europe and North America. For published in English since Jan 1, 2000. We used the search terms Asian countries, the results could also provide more insight “gastric cancer”, “stomach cancer”, “adenocarcinoma”, into the role of postoperative chemoradiotherapy than has “neoadjuvant”, “adjuvant”, “preoperative”, “postoperative”, previously been published. Finally, in view of the “chemotherapy”, “chemoradiotherapy”, “combined modality”, multidisciplinary nature of both treatment groups, we paid “gastrectomy”, “randomised”, and “phase 3”. Because of the special attention to quality assurance, ensuring that substantial difference in treatment outcome and survival state-of-the-art surgery, chemotherapy, and radiotherapy were between patients from Europe and North America versus those given to all patients (in previous studies one or more from Asia, we focused on patients from Europe and North components of treatment were suboptimal). America. In 2001, the US Intergroup 0116 trial reported that Implications of all the available evidence postoperative chemoradiotherapy improved both locoregional The results of this trial suggest that it is unlikely that patients control and overall survival compared with surgery alone, and with resectable gastric cancer treated with preoperative consequently became the standard treatment in the USA. chemotherapy and adequate surgery benefit more from In 2006, the UK MAGIC trial showed tumour downsizing and postoperative chemoradiotherapy than postoperative downstaging and improvement in survival in patients receiving chemotherapy. The implication of the results is that perioperative chemotherapy. This was corroborated by the preoperative chemotherapy with adequate surgery can be findings of the French FFCD trial in 2011, and thus made considered the backbone of resectable gastric cancer treatment. perioperative chemotherapy the preferred treatment method This trial provides a rationale to focus on preoperative in Europe. strategies and to explore further intensification of the Added value of this study preoperative phase in future studies. To our knowledge, this is the first trial directly comparing two standards of care used in high-income countries for

www.thelancet.com/oncology Vol 19 May 2018 617 Articles

788 patients enrolled and randomly assigned

393 randomly assigned to chemotherapy 395 randomly assigned to chemoradiotherapy

1 withdrew consent before treatment 6 did not start treatment 4 progressive disease 2 poor condition

392 started chemotherapy 389 started chemotherapy

58 did not receive three cycles 68 did not receive three cycles 46 toxicity 49 toxicity 4 died 4 died 3 progression or irresectability 2 progression or irresectability 3 personal event 8 personal event 1 poor condition 1 poor condition 1 withdrew consent 1 refusal 3 stomach bleeding 82 did not receive curative surgery 63 did not receive curative surgery 59 palliative resection 38 palliative resection 3 missing surgery reports* 4 missing surgery reports* 2 withdrew consent 21 no surgery 18 no surgery 12 progression or irresectability 7 progression or irresectability 4 died 6 died 3 toxicity 3 toxicity 2 refused 1 protocol deviation 1 poor condition

310 received potentially curative surgery 326 received potentially curative surgery 76 did not start postoperative 73 did not start postoperative chemotherapy chemotherapy 17 preoperative toxicity 13 preoperative toxicity 15 progression or irresectability 13 progression or irresectability 15 refused 16 refused 11 died 6 died 8 postoperative complications 18 postoperative complications 6 poor condition 4 poor condition 3 protocol deviation 3 protocol deviation 1 other 8 switched group 1 switched group due to protocol 4 refusal deviation 3 preoperative toxicity 1 protocol deviation

233 started postoperative chemotherapy 245 started postoperative chemoradiotherapy

53 did not receive three cycles 43 did not receive 5 weeks of 39 toxicity chemotherapy 4 progression or irresectability 32 toxicity 5 refusal 4 refusal 1 personal event 1 progression or irresectability 1 poor condition 1 stomach perforation 1 postoperative complication 1 postoperative complication 1 protocol deviation 4 other 1 other 24 received radiation according to protocol

5 did not receive 45 Gy of radiation (but all 5 received chemotherapy according to protocol) 2 toxicity 1 refusal 2 other 180 completed three cycles of postoperative 197 completed 5 weeks of postoperative chemotherapy chemoradiotherapy

393 included in overall survival and event-free 395 included in overall survival and event-free survival analyses survival analyses 392 included in preoperative chemotherapy safety 389 included in preoperative chemotherapy analysis safety analysis 310 included in postoperative complications and 326 included in postoperative complications pathology analyses and pathology analyses 233 included in postoperative chemotherapy 245 included in postoperative chemoradiotherapy safety analysis safety analysis

618 www.thelancet.com/oncology Vol 19 May 2018 Articles

Patients were eligible if they had histologically proven the diagnostic process, and to avoid patient selection bias stage IB–IVA gastric adenocarcinoma (as defined by the after surgery. Patients were enrolled and treated by the American Joint Committee on Cancer, sixth edition),9 as local investigators or their delegated staff at participating assessed by oesophagogastroduodenoscopy and CT of the centres, and registration of eligible patients could be done chest, abdomen, and pelvis. Patients with tumours of the by telephone, fax, or online for the Central Data center of gastro-oesophageal junction were permitted to enrol the Department of Surgery at Leiden University Medical when the bulk of the tumour was predominantly located Center. After verification of eligibility criteria, we used the in the stomach, and could therefore consist of Siewert ALEA Randomisation computer programme, which types II and III tumours. Because the focus of the study implements a minimisation technique described by was on gastric cancer, patients with Siewert type I tumours Pocock and Simon,10 to randomly assign patients and were not eligible. A diagnostic laparoscopy was indicated stratify them according to histological subtype (Lauren when the preoperative CT scan suggested peritoneal classification: intestinal vs diffuse vs mixed vs unknown), carcinomatosis. PET scans were optional and were done tumour location (gastro-oesophageal junction vs proximal according to local practice when clinically indicated. To be stomach vs middle stomach vs distal stomach), and eligible, patients also had to be 18 years or older, have a hospital. Patients and investigators were not masked to WHO performance status of 0 or 1, have adequate cardiac, treatment allocation. bone marrow, liver, and kidney function, and have had no previous radiotherapy or chemotherapy that would affect Procedures treatment for gastric cancer. Eligible participants also had Preoperative chemotherapy consisted of three 21-day a left ventricular ejection fraction of at least 50% and cycles of epirubicin, cisplatin or oxaliplatin, and urinary protein excretion of less than 1 g per 24 h. In case capecitabine. Epirubicin 50 mg/m², cisplatin 60 mg/m², of insufficient caloric intake or substantial weight loss, and oxaliplatin 130 mg/m² were given intravenously on oral nutritional support or enteral tube feeding was day 1 of each 21-day cycle. Capecitabine was administered warranted. Exclusion criteria included T1N0 tumours as assessed by endoscopic ultrasound and previous malignancy, except adequately treated non-melanoma Chemotherapy Chemoradiotherapy skin cancer and in-situ cancer of the cervix uteri. Other group (n=393) group (n=395) reasons for exclusion were a solitary functioning kidney Age (years) that would be located within the radiation field, major Median age (IQR) 62 (54–69) 63 (56–68) surgery within 4 weeks before start of study treatment, or <60 164 (42%) 155 (39%) lack of complete recovery from previous surgery, 60–69 142 (36%) 155 (39%) uncontrolled cardiac or infectious disorders, continuous ≥70 87 (22%) 85 (22%) use of immunosuppressive drugs, or other conditions Sex preventing the safe use of study drugs and treatment Male 264 (67%) 265 (67%) methods. A full list of exclusion criteria is provided in the Female 129 (33%) 130 (33%) appendix (p 15). On the basis of the survival curve of the WHO performance status perioperative chemotherapy group in the MAGIC study,6 0 260 (66%) 274 (69%) which showed a median overall survival of approximately 1 103 (26%) 106 (27%) 2 years, we estimated that life expectancy of patients in Unknown 30 (8%) 15 (4%) this study would be about 2 years. Histological subtype* The study was approved by the medical ethical Intestinal 127 (32%) 126 (32%) committee of the Netherlands Cancer Institute and by Diffuse 116 (30%) 117 (30%) the review boards of all participating centres. All patients Mixed 20 (5%) 22 (6%) provided oral and written informed consent. Unknown 130 (33%) 130 (33%) Tumour localisation Randomisation and masking Gastro-oesophageal junction 68 (17%) 67 (17%) Patients were randomly assigned (1:1) to either Proximal stomach 79 (20%) 84 (21%) perioperative chemotherapy or preoperative chemo Middle stomach 120 (31%) 117 (30%) therapy and postoperative chemoradiotherapy. We chose Distal stomach 126 (32%) 127 (32%) to do randomisation before the start of preoperative Diagnostic laparoscopy chemotherapy because this approach most closely reflects Done 36 (9%) 43 (11%) daily practice, with treatment decisions being made after Not done 356 (91%) 352 (89%) Unknown 1 (<1%) 0

Figure 1: Trial profile *Histological subtype according to the Lauren classification. *Of the seven patients with missing surgery reports, six received postoperative Table 1: Baseline characteristics treatment as planned and one patient had no postoperative treatment.

www.thelancet.com/oncology Vol 19 May 2018 619 Articles

at doses of 1000 mg/m² orally two times per day as tablets Each patient was assessed for medical history, physical for 14 days in the epirubicin, cisplatin, and capecitabine examination, and blood tests before each chemotherapy regimen, and 625 mg/m² orally twice daily for 21 days in cycle and once per week during chemoradiotherapy. We the epirubicin, oxaliplatin, and capecitabine regimen. In assessed toxicity before each chemotherapy cycle, once case of difficulties with ingestion of tablets, capecitabine per week during chemoradiotherapy, and once per month could be replaced by fluorouracil 200 mg/m² daily by during follow-up until 3 months, according to National continuous infusion for 21 days. Response assessment Cancer Institute Common Toxicity Criteria for Adverse with CT scan was done after two chemotherapy cycles to Events (CTCAE; version 3.0). Dose reductions in chemo exclude early progression. When preoperative chemo therapy were permitted according to guidance in the study therapy was postponed for more than 2 consecutive protocol, with a maximum dose reduction of 50% for weeks, chemotherapy was discontinued and the patient capecitabine in case of persisting grade 3 or 4 adverse proceeded to surgery when possible. events. Cisplatin or oxaliplatin were discontinued in The intent of surgery was a radical resection of the patients developing significant nephrotoxicity, ototoxicity, primary tumour (by means of a total gastrectomy, subtotal or sensory neuro-toxicity during preoperative and post gastrectomy, or oesophagocardiac resection) en bloc with operative treatment. We categorised postoperative the N1 and N2 lymph nodes (stations 1–9 and 11) and complications into general (eg, cardiovascular, pulmonary, a minimum of 15 lymph nodes (D1+ lymph node renal, and neurological complications), infectious dissection),11 and, if possible, a macroscopic proximal and (eg, abdominal wound, abscess, and sepsis), and surgery- distal margin of 5 cm. A potentially curative resection was related complications (bleeding, anastomotic leakage, defined as no evidence of macroscopic residual disease at abdominal wound dehiscence, ileus, and intestinal the end of the operation, as judged by the surgeon. necrosis). After postoperative treatment patients had Postoperative treatment had to start within 4–12 weeks follow-up visits every month during the first 3 months, after surgery. The postoperative chemotherapy regimen every 3 months during the rest of the first year, followed consisted of the same chemotherapy regimen as by every 6 months until 5 years. During these follow-up administered preoperatively. Postoperative chemo- visits, history, physical examination, and blood tests were radiotherapy was based on previous dose-finding assessed. CT scans of the thorax, abdomen, and pelvis studies.12–14 Capecitabine was administered at a dose of were done every 6 months during the first 2 years and 575 mg/m² orally twice daily on radiotherapy days, for 5 then once per year until 5 years. During follow-up, weeks, with five daily fractions per week. Cisplatin was renography was done once per year in the chemo administered at a dose of 20 mg/m² intravenously on radiotherapy group. the first day of each 5 weeks of radiotherapy treatment. On-site data monitoring was done with source The radiation dose was 45 Gy, given in 25 fractions of verification for informed consent, inclusion and 1·8 Gy, for 5 weeks, with five daily fractions per week. exclusion criteria, protocol procedures, missing data, and The clinical target volume included the tumour bed, serious adverse events for at least the first five registered surgical anastomoses, and regional lymph node stations, patients in all participating centres. Surgical and which were individualised according to tumour pathological quality assurance involved central review for localisation. The clinical target volume had to be type and completeness of resection, including number of delineated on CT images. lymph nodes retrieved. Radiotherapy quality assurance consisted of pre-treatment assessment of treatment plans of at least the first three patients included by each Chemotherapy Chemoradiotherapy radiotherapy institute, and for subsequent patients if group group deemed necessary by the principle investigators Preoperative chemotherapy n=392 n=389 (EPMJ and MV), or requested by the treating radiation Epirubicin 97% (82–100) 95% (76–100) oncologist. Target volume delineation manuals and Cisplatin 95% (81–100) 96% (82–100) workshops were offered to all participating institutions. Oxaliplatin 98% (87–100) 93% (85–99) Capecitabine 91% (73–100) 90% (71–100) Outcomes Postoperative chemotherapy n=233 n=245 The primary endpoint was overall survival, defined as the Epirubicin* 83% (63–99) .. time from randomisation to death from any cause. Cisplatin 87% (66–99) 98% (80–100) Secondary endpoints were event-free survival, toxicity, and Oxaliplatin* 72% (59–97) .. health-related quality of life. Other prespecified exploratory Capecitabine 76% (54–93) 90% (77–100) secondary endpoints were the prediction of response and recurrence risk by genomic and proteomic profiling, and Values are median percentages (IQR) of recommended dose intensities. *Patients who received postoperative chemoradiotherapy did not receive epirubicin and determination of the value of Maruyama Index and oxaliplatin during radiotherapy. predictive nomograms for disease recurrence. Event-free survival was defined as time from ran-domisation until Table 2: Administered doses of chemotherapeutic drugs disease progression, irresectable disease at surgery,

620 www.thelancet.com/oncology Vol 19 May 2018 Articles

locoregional or peritoneal tumour recurrence, distant would be lost to follow-up, an estimated sample size of metastases, or death from any cause, whichever occurred 780 patients would be sufficient (after protocol first. Irresectable disease was assessed centrally on the amendment on Jan 9, 2007). One interim analysis was basis of surgery and pathology reports, and the other planned after half of the events had been observed. The events were assessed by the local investigator. Recurrences rejection boundaries were set using the O’Brien-Flemming occurring at different sites within a period of 30 days were method. We did the interim analysis in April 8, 2014, after scored as multiple sites of recurrence. Because the health- 254 events. The independent data monitoring committee related quality of life outcome, Maruyama Index, and concluded that accrual could continue as planned. The predictive nomograms represent extensive and distinct final analyses are based on data received until July 14, 2017. topics, these outcomes will be reported separately. The Efficacy analyses were based on the intention-to-treat analyses of genomic and proteomic profiles are ongoing principle and included all randomly assigned patients. and will be based on the biomaterial available, because For overall survival and event-free survival, patients for this was an optional item at the patients’ choice. whom no death or event was observed were censored on the date last seen by the physician. Safety data were Statistical analysis analysed separately for preoperative and postoperative Based on results of the Intergroup 0116 and MAGIC periods and included patients who received at least one trials, we expected 5-year overall survival in the chemotherapy group to be 40%. We expected the Chemotherapy Chemoradiotherapy chemoradiotherapy regimen to increase 5-year overall group (n=310) group (n=326) survival to 50% in this group. To achieve 80% power to (Continued from previous column) detect this effect at an α level of 5%, 405 events were ypT stage* needed (after protocol amendment on Jan 9, 2007). pT0 20 (6%) 20 (6%) Assuming that accrual would take 4 years, the last patient pTis 1 (<1%) 5 (2%) would be followed up for 3 years, and 2% of patients pT1 41 (13%) 46 (14%) pT2a 54 (17%) 58 (18%)

Chemotherapy Chemoradiotherapy pT2b 54 (17%) 56 (17%) group (n=310) group (n=326) pT3 110 (35%) 107 (33%) Type of resection pT4 30 (10%) 34 (10%) Oesophagocardiac resection 32 (10%) 31 (10%) ypN stage* Subtotal gastrectomy 119 (38%) 136 (42%) pN0 150 (48%) 161 (49%) Total gastrectomy 159 (51%) 159 (49%) pN1 109 (35%) 105 (32%) Type of lymph node dissection pN2 35 (11%) 42 (13%)

www.thelancet.com/oncology Vol 19 May 2018 621 Articles

dose of the corresponding treatment. The survival administration in terms of relative dose intensity. curves were constructed using the Kaplan-Meier We calculated the relative dose intensities were method and compared using the two-sided log-rank test calculated for each drug according to a formula, wherein stratified for histological subtype and tumour the dose planned is calculated according to the patients’ localisation. We calculated hazard ratios (HRs) using a square metre body surface area) at the start of the stratified proportional-hazards Cox model. We assessed preoperative or postoperative treatment: proportionality of hazards using the Grambsch-Therneau 15 test. We tested the homogeneity of the treatment effect dose received (mg) Relative dose intensities = × across the levels of baseline factors (the stratification dose planned (mg) factors and post-hoc analyses for age and sex) using the time planned (days) interaction term in a Cox model, and displayed the × 100% results in a forest plot. In the subgroup analyses, we time planned + days of delay present 99% CIs for the HRs to account for additional uncertainty due to multiple testing. We summarised We did all analyses using R (version 3.3.1). dose reductions and delays in chemotherapeutic drug The CRITICS trial is registered at ClinicalTrials.gov, number NCT00407186; EudraCT, number 2006-004130-32; and CKTO 2006-02. A 100 Chemotherapy group Role of the funding source Chemoradiotherapy group The funders of the study had no role in study design,

80 data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final 60 responsibility for the decision to submit for publication.

Results 40 Between Jan 11, 2007, and April 17, 2015, we enrolled Overall survival (%) 788 patients and randomly assigned them to receive postoperative chemotherapy (n=393) or postoperative 20 chemoradiotherapy (n=395; figure 1). Seven patients did not start treatment but were included in efficacy analyses HR 1·01 (95% CI 0·84–1·22); p=0·90 0 (figure 1). Patient demographics and baseline disease 0 9684726048362412 characteristics were well balanced between the two Number at risk (number censored) groups (table 1). Chemotherapy group 393(0) 298 (0) 230 (6) 176 (31) 127(69) 88 (94) 58 (120) 36 (142) 23 (154) Chemoradiotherapy group 395(0) 298 (1) 237 (4) 176 (25) 123(62) 84 (89) 58 (111) 35 (132) 17 (149) Median time from randomisation to start of preoperative chemotherapy was 2 days (IQR 1–4) in the chemotherapy B group and 2 days (1–5) in the chemo-radiotherapy group. 100 In 15 (2%) of the 781 preoperatively treated patients (8 in the chemotherapy group and 7 in the chemoradiotherapy

80 group), infusional fluorouracil was given because of ingestion problems, and 149 (19%) of 781 preoperatively treated patients received oxaliplatin (78 in the chemotherapy 60 group and 71 in the chemoradiotherapy group). In the 392 patients who actually received preoperative chemotherapy in the postoperative chemotherapy group, 17 (4%) patients 40 received one cycle of preoperative chemotherapy, 41 (10%)

Event-free survival (%) Event-free received two cycles, and 334 (85%) received three cycles. In the chemoradiotherapy group (n=389), 27 (7%) of 20 389 patients received one cycle of preoperative chemotherapy, 41 (11%) received two cycles, and 321 (83%) HR 0·99 (95% CI 0·82–1·19); p=0·92 0 received three cycles. Median percentages of recommended 0 12 24 36 48 60 72 84 96 dose intensities for all administered chemotherapeutic Time since randomisation (months) Number at risk (number censored) drugs are shown in table 2. Chemotherapy group 393(0) 249 (0) 199(5) 153 (27) 112(61) 81 (83) 55 (107) 34 (127) 22 (138) After preoperative chemotherapy, 372 (95%) of Chemoradiotherapy group 395(0) 254 (1) 197 (4) 159 (21) 114(56) 81 (80) 55 (101) 33 (122) 17 (138) 393 patients in the chemotherapy group and 369 (93%) of Figure 2: Overall survival (A) and event-free survival (B) 395 patients in the chemoradiotherapy group proceeded HR=hazard ratio. to surgery. Surgery was done after a median of 4 weeks

622 www.thelancet.com/oncology Vol 19 May 2018 Articles

(IQR 2–5) since the last chemotherapy cycle in the 221 (90%) of 245 patients in the chemoradiotherapy group. chemotherapy group and 4 weeks (3–5) in the Of 105 patients who did not have potentially curative chemoradiotherapy group. A potentially curative resection, surgery or started the allocated treatment, 65 (62%) confirmed by assessment of the surgery report, was done patients started another postoperative treatment outside in 310 (79%) of 393 patients in the chemotherapy group the trial protocol: 31 patients in the chemotherapy group versus 326 (83%) of 395 patients in the chemoradiotherapy and 34 patients in the chemoradiotherapy group. group (table 3). The type of resection and lymph node After a median follow-up of 61·4 months dissection were well balanced between both groups, as (IQR 43·3–82·8), 216 (55%) of 393 patients in the was the number of harvested lymph nodes (table 3). In chemotherapy group and 230 (58%) of 395 patients in the 544 (86%) of these 636 patients, at least a D1+ resection chemoradiotherapy group had died. Overall survival was was done (table 3). not significantly different between the groups according to A microscopically incomplete resection (R1) was the intention-to-treat analysis (figure 2A). Median overall documented in 66 (10%) of 636 patients who had a survival was 43 months (95% CI 31–57) in the chemotherapy curative resection, as judged by the surgeon. 37 (6%) group and 37 months (30–48) in the chemoradiotherapy patients in this group achieved a pathological complete group (hazard ratio [HR] from stratified analysis 1·01 response. Tumour and lymph node stage were equally [95 CI% 0·84–1·22]; p=0·90). 5-year survival probabilities distributed in both groups, as were stage groups (table 3). were 42% (95% CI 37–48) for chemotherapy and After potentially curative surgery, 233 patients in the 40% (35–46) for chemoradiotherapy. chemotherapy group started postoperative treatment after For the analysis of event-free survival, 474 events were a median of 7 weeks (IQR 6–9), and 245 patients in the recorded in 788 patients overall (233 in the chemotherapy chemoradiotherapy group started postoperative treatment group vs 241 in the chemoradiotherapy group). Median after 8 weeks (7–10) weeks. One patient in the chemotherapy event-free survival was 28 months (95% CI 20–42) in the group and eight patients in the chemoradiotherapy group chemotherapy group versus 25 months (19–39) in the switched their treatment to that of the other group chemoradiotherapy group (HR from stratified analysis (two because of protocol deviations, four because of 0·99, 95% CI 0·82–1·19, p=0·92), and 5-year event-free treatment refusals, and three because of toxicities). survival probabilities were 39% (34–44) versus 38% (33–44; Reasons for not starting postoperative treatment were figure 2B). The event-free survival events were locoregional death in 17 (4%) of 393 patients in the chemotherapy group recurrence (35 [15%] of 233 in the chemotherapy group vs and 10 (3%) of 392 patients in the chemoradiotherapy 27 [11%] of 241 in the chemoradiotherapy group), group, progressive or irresectable disease in 81 (21%) and peritoneal recurrence (50 [21%] vs 61 [25%]), distant 67 (17%; including patients with progressive disease before recurrence (58 [25%] vs 52 [22%]), and recurrence at the start of chemoradiotherapy, palliative resection, multiple sites (50 [21%] vs 65 [27%]). 76 (16%) of all progressive disease in the no surgery group, and previous preoperative toxicity); treatment-related toxicity (ie, toxicity Chemotherapy Chemoradiotherapy Hazard ratio pinteraction group (n/N) group (n/N) (99% CI) in patients who had no surgery, previous properative toxicity, and postoperative complications) in 28 (7%) and Histology 0·236 Intestinal 54/127 68/126 1·35 (0·85–2·17) 34 (9%); and refusal or poor general health such that Diﬀuse 76/116 74/117 0·88 (0·58–1·35) continuation of treatment was not possible in 22 (6%) and Mixed 8/20 12/22 1·46 (0·44–4·78) Unknown 78/130 76/130 0·93 (0·62–1·41) 24 patients (6%; including patients with poor condition Location 0·163 before start of preoperative chemotherapy, those who Gastro-oesophageal 34/68 38/67 1·12 (0·61–2·06) refused in the no surgery group, those who refused after junction Proximal 44/79 56/84 1·33 (0·79–2·24) surgery, and those with poor condition after surgery; Middle 65/120 73/117 1·17 (0·75–1·81) figure 1). Of the patients who started postoperative Distal 73/126 63/127 0·77 (0·49–1·19) Sex 0·0012 treatment, 16 (7%) of 233 patients in the chemotherapy Male 135/264 163/265 1·30 (0·96–1·76) group received one cycle, 37 (16%) received two cycles, and Female 81/129 67/130 0·68 (0·44–1·04) 180 (77%) received all three cycles of postoperative Age (years) 0·065 <60 93/164 88/155 0·94 (0·64–1·38) chemotherapy. In the chemoradiotherapy group, of the 60–69 69/142 94/155 1·40 (0·93–2·10) 245 patients who started postoperative chemo-radiotherapy, ≥70 54/87 48/85 0·81 (0·48–1·35) six (2%) patients received 1 week of postoperative Overall 216/393 230/395 1·01 (0·84–1·22) chemotherapy, eight (3%) patients received 2 weeks, 0·5 0·75 1 1·5 2 ten (4%) patients received 3 weeks, 19 (18%) patients Favours Favours received 4 weeks, and 202 (82%) patients received all chemoradiotherapy chemotherapy 5 weeks (figure 1). In the intention-to-treat population, 180 (46%) of 393 patients in the chemotherapy group and Figure 3: Subgroup analyses of overall survival The effect of trial treatment on overall survival according to baseline characteristics, including histological subtype 197 (50%) of 395 patients in the chemoradiotherapy according to Lauren classification, tumour location, sex, and age. Hazard ratios for subgroups come from Cox group completed the allocated treatment as planned. models corrected only for treatment group, and that for the overall result comes from a Cox model adjusted for Radiotherapy could be administered as planned in treatment group, histology, and location. n=number of events. N=number of patients. www.thelancet.com/oncology Vol 19 May 2018 623 Articles

474 events were deaths without known progression, events during preoperative treatment were reported in irresectability, or recurrence (40 [17%] of 233 in the 767 (98%) of 781 patients who started preoperative chemotherapy group vs 36 [15%] of 241 in the treatment (the safety population): 52 (7%) had chemoradiotherapy group). There was no evidence that grade 1 events, 204 (26%) had grade 2 events, the assumption of proportional hazards was violated. 368 (47%) had grade 3 events, 130 (17%) had grade 4 events, In a post-hoc subgroup analysis of overall survival, and 13 (2%) had grade 5 events (table 4). Causes of death there was no clear evidence for heterogeneity of treatment were diarrhoea (n=2), dihydropyrimidine deficiency effect according to age, histology, and location of the (n=1), sudden death (n=1), cardiovascular events (n=8; tumour at the time of randomisation, but there was for two cardiac arrhythmia and six ischaemic events), and sex (figure 3). gastrointestinal obstruction (n=1). All but one Because preoperative treatment was identical in both cardiovascular event were considered to be treatment groups, toxicity scoring during this period was done in related. Drug-related serious adverse events were both groups together (all treated patients). Adverse reported in 227 (29%) of 781 patients: 176 (23%) had a

Preoperative chemotherapy (n=781) Postoperative chemotherapy (n=233) Postoperative chemoradiotherapy (n=245) Grade 1–2 Grade 3 Grade 4 Grade 5 Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 Haematological Anaemia 448 (57%) 18 (2%) 6 (1%) 0 145 (62%) 1 (<1%) 0 141 (58%) 2 (1%) 0 Leucopenia 60 (8%) 14 (2%) 7 (1%) 0 13 (6%) 1 (<1%) 0 21 (9%) 3 (1%) 0 Neutropenia 220 (28%) 174 (22%) 76 (10%) 0 66 (28%) 62 (27%) 17 (7%) 23 (9%) 10 (4%) 1 (<1%) Lymphocytopenia 0 0 0 0 0 1 (<1%) 0 1 (<1%) 2 (1%) 0 Febrile neutropenia 0 54 (7%) 10 (1%) 0 0 4 (2%) 1 (<1%) 0 5 (2%) 0 Thrombocytopenia 39 (5%) 3 (<1%) 2 (<1%) 0 4 (2%) 0 1 (<1%) 40 (16%) 3 (1%) 1 (<1%) Gastrointestinal Mucositis or stomatitis 235 (30%) 30 (4%) 2 (<1%) 0 43 (18%) 6 (3%) 0 29 (12%) 2 (1%) 0 Heartburn or dyspepsia 53 (7%) 2 (<1%) 0 0 20 (9%) 4 (2%) 0 39 (16%) 5 (2%) 0 Dysphagia 72 (9%) 19 (2%) 0 0 26 (11%) 5 (2%) 0 37 (15%) 7 (3%) 1 (<1%) Anorexia 288 (37%) 71 (9%) 2 (<1%) 0 84 (36%) 20 (9%) 0 111 (45%) 35 (14%) 0 Nausea 459 (59%) 83 (11%) 1 (<1%) 0 130 (56%) 27 (12%) 0 171 (70%) 23 (9%) 0 Vomiting 251 (32%) 58 (7%) 3 (<1%) 0 84 (36%) 10 (4%) 0 92 (38%) 12 (5%) 0 Diarrhoea 226 (29%) 95 (12%) 5 (1%) 2 (<1%) 86 (37%) 13 (6%) 0 82 (33%) 8 (3%) 0 Constipation 265 (34%) 4 (1%) 1 (<1%) 0 40 (17%) 3 (1%) 0 59 (24%) 0 0 Bowel inflammation 2 (<1%) 6 (1%) 2 (<1%) 0 1 (<1%) 2 (1%) 0 0 1 (<1%) 0 Gastrointestinal fistula 0 0 0 0 1 (<1%) 0 0 0 1 (<1%) 0 Gastrointestinal obstruction 7 (1%) 7 (1%) 2 (<1%) 1 (<1%) 2 (1%) 2 (1%) 0 0 2 (1%) 0 Gastrointestinal perforation 1 (<1%) 4 (1%) 0 0 0 1 (<1%) 0 0 0 2 (1%) Vascular Cardiac arrhythmia 21 (3%) 4 (1%) 0 2 (<1%) 3 (1%) 0 0 7 (3%) 0 0 Ischaemic event 5 (1%) 14 (2%) 7 (1%) 6 (1%) 0 2 (1%) 1 (<1%) 0 1 (<1%) 0 Thromboembolic event* 24 (3%) 44 (6%) 21 (3%) 0 0 3 (1%) 2 (1%) 0 2 (1%) 1 (<1%) Sudden death 0 0 0 1 (<1%) 0 0 0 0 0 0 Haemorrhage 38 (5%) 8 (1%) 1 (<1%) 0 5 (2%) 1 (<1%) 0 5 (2%) 0 0 Constitutional Weight loss 191 (24%) 10 (1%) 0 0 77 (33%) 5 (2%) 0 99 (40%) 2 (1%) 0 Dehydration 34 (4%) 42 (5%) 2 (<1%) 0 4 (2%) 3 (1%) 0 9 (4%) 8 (3%) 0 Dizziness 63 (8%) 17 (2%) 0 0 19 (8%) 4 (2%) 0 16 (7%) 3 (1%) 0 Fatigue 483 (62%) 57 (7%) 8 (1%) 0 152 (65%) 20 (9%) 0 174 (71%) 27 (11%) 0 Hypertension 23 (3%) 12 (2%) 0 0 3 (1%) 0 0 5 (2%) 3 (1%) 0 Infection without neutropenia 115 (15%) 59 (8%) 5 (1%) 3 (<1%) 26 (11%) 8 (3%) 2 (1%) 31 (13%) 12 (5%) 2 (1%) Insomnia 16 (2%) 1 (<1%) 1 (<1%) 0 7 (3%) 0 0 10 (4%) 1 (<1%) 0 Mood alteration 62 (8%) 5 (1%) 1 (<1%) 0 24 (10%) 1 (<1%) 0 22 (9%) 1 (<1%) 0 Pain, abdominal 109 (14%) 8 (1%) 0 0 34 (15%) 4 (2%) 0 52 (21%) 4 (2%) 0 Pain, other 150 (19%) 25 (3%) 1 (<1%) 0 30 (13%) 8 (3%) 1 (<1%) 32 (13%) 10 (4%) 0 (Table 4 continues on next page)

624 www.thelancet.com/oncology Vol 19 May 2018 Articles

Preoperative chemotherapy (n=781) Postoperative chemotherapy (n=233) Postoperative chemoradiotherapy (n=245) Grade 1–2 Grade 3 Grade 4 Grade 5 Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 (Continued from previous page) Other Allergic reaction 21 (3%) 5 (1%) 0 0 0 0 0 2 (1%) 0 0 Alopecia 317 (41%) 0 0 0 64 (27%) 0 0 35 (14%) 0 0 Dermatological reaction 121 (15%) 3 (<1%) 0 0 21 (9%) 0 0 23 (9%) 2 (1%) 0 Dihydropyrimidine dehydrogenase 0 0 0 1 (<1%) 0 0 0 0 0 0 deficiency Dyspnoea 77 (10%) 2 (<1%) 0 0 25 (11%) 0 0 21 (9%) 1 (<1%) 0 Genitourinary obstruction 0 0 1 (<1%) 0 0 0 0 0 0 0 Local complication (device related) 53 (7%) 5 (1%) 0 0 13 (6%) 2 (1%) 0 2 (1%) 4 (2%) 0 Metabolic disorder† 65 (8%) 56 (7%) 16 (2%) 0 24 (10%) 5 (2%) 1 (<1%) 37 (15%) 11 (4%) 2 (1%) Musculoskeletal disorder 31 (4%) 3 (<1%) 1 (<1%) 0 5 (2%) 0 0 9 (4%) 0 0 Neurotoxicity 230 (29%) 9 (1%) 2 (<1%) 0 83 (36%) 4 (2%) 0 43 (18%) 1 (<1%) 0 Ototoxicity 48 (6%) 1 (<1%) 0 0 22 (9%) 0 0 14 (6%) 1 (<1%) 0 Palmar-plantar erythrodysesthesia 204 (26%) 23 (3%) 1 (<1%) 0 44 (19%) 3 (1%) 0 16 (7%) 0 0 Psychosis 0 2 (<1%) 0 0 0 0 0 0 0 0 Renal toxicity 49 (6%) 11 (1%) 1 (<1%) 0 5 (2%) 3 (1%) 0 12 (5%) 2 (1%) 0 Other toxicity 92 (12%) 3 (<1%) 0 0 29 (12%) 0 0 31 (13%) 0 0 Any adverse event 256 (33%) 368 (47%) 130 (17%) 13 (2%) 94 (40%) 113 (48%) 22 (9%) 128 (52%) 101 (41%) 10 (4%)

We scored adverse events according to the National Cancer Institute Common Toxicity Criteria for Adverse Events (version 3.0). All adverse events that occurred as grade 3–5, or adverse events which occurred only as grade 1–2 in at least 10% of the population, are reported. There were no postoperative grade 5 events. We scored toxicity during the preoperative treatment period from the first preoperative chemotherapy cycle until the last preoperative chemotherapy cycle administered plus 30 days, or surgery, whatever occurred first. Toxicity during the postoperative treatment period was scored from the first postoperative treatment method (either chemotherapy or radiotherapy) until the last postoperative treatment method administered plus 30 days. *Also includes patients with superficial phlebitis. †Worsening kidney or liver functioning, or both (measured by laboratory tests).

Table 4: Adverse events associated with preoperative and postoperative treatment grade 3 serious adverse event, 39 (5%) had a grade 4 event, treatment, no grade 5 adverse events were reported. and 12 (2%) had a grade 5 event. Of these serious adverse Grade 3–4 non-febrile neutropenia occurred more events, diarrhoea (64 [8%]), febrile neutropenia (54 [6%]), frequently during postoperative chemotherapy (79 [34%] nausea (34 [4%]), vomiting (34 [4%]), and dehydration of 233) than during postoperative chemoradiotherapy (28 [4%]) were the most frequently reported of these (11 [4%] of 245). During postoperative chemotherapy, 227 serious adverse events. treatment-related serious adverse events were reported General, infectious, and surgery-related complications in 37 (16%) of 233 patients in the chemotherapy group during and after surgery occurred each in 22–28% of versus 39 (16%) of 245 in the chemoradiotherapy group. patients who had potentially curative surgery, and did not The most frequently reported serious adverse events in differ between groups (table 5). In these patients (n=636), the chemotherapy group were nausea (13 [6%]), diarrhoea surgical morbidity consisted of anastomotic leakage (11 [5%]), and vomiting (six [3%]). The most frequently (45 [7%] overall; 18 [6%] in the chemotherapy group vs reported serious adverse events in the chemoradiotherapy 27 [8%] in the chemoradiotherapy group), ileus (18 [3%]; group were fatigue (ten [4%]) and nausea (eight [3%]). ten [3%] vs eight [2%]), bleeding (18 [3%]; ten [3%] vs eight [2%]), abdominal wound dehiscence (ten [2%]; Discussion three [1%] vs seven [2%]), fistulae (six [1%]; four [1%] vs The findings of this study do not support the hypothesis two [1%]), and other (73 [11%]; 37 [12%] vs 36 [11%]). that treatment consisting of preoperative chemotherapy, 14 (2%) of 636 patients died in the hospital or within surgery, and postoperative chemoradiotherapy improve 30 days after surgery (ten in the chemotherapy group and survival compared with perioperative chemotherapy and four in the chemoradiotherapy group). surgery in patients with resectable gastric and gastro- During postoperative treatment, 16 (7%) of 233 patients oesophageal adenocarcinoma. This randomised study had grade 1 adverse events in the postoperative chemo integrated two treatment strategies of perioperative therapy group versus 30 (12%) of 245 in the postoperative chemotherapy and postoperative chemoradiotherapy that chemoradiotherapy group; 78 (33%) versus 98 (40%) had previously showed beneficial outcomes in patients from grade 2 adverse events; 113 (48%) versus 101 (41%) had Europe and North America. grade 3 adverse events; and 22 (9%) versus ten (4%) had Randomisation before the start of any treatment grade 4 adverse events (table 4). During postoperative prevented patient selection during the treatment process, www.thelancet.com/oncology Vol 19 May 2018 625 Articles

oesophagus or gastro-oesophageal junction.20 In the Chemotherapy group (n=310) Chemoradiotherapy group (n=326) CALGB 80101 study,21 the addition of epirubicin and General complications* cisplatin to postoperative chemotherapy in combination None 223 (72%) 231 (71%) with chemoradiotherapy also did not improve survival in Yes 86 (28%) 94 (29%) patients with gastric or gastro-oesophageal Unknown 1 (<1%) 1 (<1%) adenocarcinomas. Results from the FLOT4 study19 showed Infectious complications† that perioperative chemotherapy with docetaxel, None 235 (76%) 251 (77%) oxaliplatin, fluorouracil, and leucovorin increased both the Yes 74 (24%) 74 (23%) proportion of patients who achieved a pathological Unknown 1 (<1%) 1 (<1%) complete response and overall survival compared with Surgery-related complications‡ perioperative epirubicin, cisplatin, and fluorouracil or None 239 (77%) 253 (78%) capecitabine in patients with gastric or gastro-oesophageal Yes 70 (23%) 72 (22%) adenocarcinomas. Overall, the potential downsizing and Unknown 1 (<1%) 1 (<1%) downstaging effects of preoperative chemotherapy and the Reintervention for the management of complications limitations of the amount of treatment that can be No 269 (87%) 276 (85%) administered postoperatively to patients with gastric and Yes 37 (12%) 48 (15%) gastro-oesophageal cancer implies that additional survival Unknown 4 (1%) 2 (1%) benefit is most likely to come from optimisation of preoperative treatment strategies. The complications listed in this table were recorded during hospital stay after surgery. *Including cardiovascular, pulmonary, renal, and neurological complications. †Including abdominal wound, abscess, and sepsis. ‡Including Subsequently, the question arises about whether or bleeding, anastomotic leakage, abdominal wound dehiscence, ileus, and intestinal necrosis. when chemoradiotherapy adds to the efficacy of preoperative or postoperative chemotherapy. It is possible Table 5: Surgical complications in patients who had potentially curative surgery that because all patients in our study received preoperative chemotherapy, postoperative treatment did not further and thus most closely reflected daily practice and feasibility increase survival. Additionally, the percentage of observed of the entire treatment regimen in both groups. With 94% events that attributed exclusively to locoregional of patients proceeding to surgery and receiving recurrence was similar and low in both groups. Results recommended dose intensities of more than 90% for all from the Intergroup study4 previously established the preoperatively administered chemotherapy drugs, the survival benefit of postoperative fluorouracil monotherapy results of our study show that preoperative chemotherapy in combination with fluorouracil-based chemoradiotherapy is feasible. Around 50% of patients completed treatment versus surgery only in medically fit patients who had an as planned, which compares slightly favourably with other R0 resection. An observational study22 from South Korea, studies investigating perioperative chemotherapy.6,16–19 which compared postoperative chemoradiotherapy Nevertheless, postoperative treatment was hampered by according to the Intergroup protocol with surgery only substantial patient dropout, mainly due to irresectability (but including a formal D2 lymph node dissection), of the primary tumour or early progression after surgery showed similar results—namely, that overall survival (approximately 18% of patients who had palliative improved with postoperative chemoradiotherapy.22 By resection plus those with progressive disease), and contrast, the subsequent randomised ARTIST trial23 treatment-related toxicity, death, worsening of general compared adjuvant chemotherapy (capecitabine plus health that precluded continuation of treatment (almost cisplatin) with or without capecitabine-based chemo 20%), or patients’ refusal to continue treatment radiotherapy in patients who had D2-resected gastric (nearly 20% of patients). Additionally, the doses of cancer, but reported no difference in disease-free survival. administered postoperative chemotherapeutic drugs Several factors, such as study period, ethnicity, surgical decreased to below 80–90% of the recommended dose quality, and the addition of systemic chemotherapy to the intensities, which emphasises the limitations of the extent control group might account for these conflicting findings. of treatment that can be given to this patient population. Either treatment method might contribute to improving Results from the MAGIC study6 of perioperative survival for as long as it is administered, or alternatively, chemotherapy versus surgery alone for resectable some patient subgroups might benefit from one of these gastroesophageal cancer showed that preoperative treatments. epirubicin, cisplatin, and fluorouracil leads to tumour Surgical quality in this study was good, with a median downsizing and downstaging, and also to enhanced overall lymph node yield of 20, an R0 resection rate of at least survival versus surgery alone. However, the contribution of 81%, and a Maruyama Index (reported separately)24 of 1. epirubicin to this favourable effect could be questioned. In A Cochrane systematic review and meta-analysis25 of the phase 3 UK MRC OE05 trial, four cycles of preoperative eight randomised controlled trials consisting of epirubicin, cisplatin, and capecitabine compared with 2515 patients has advocated D2 dissection over two cycles of cisplatin and fluorouracil did not increase D1 and D3 dissections on the basis of a better disease- survival in patients with adenocarcinomas of the specific survival.25 Because of the increased postoperative

626 www.thelancet.com/oncology Vol 19 May 2018 Articles

mortality ascribed to the splenic and pancreatic overexpression, MSI, EBV, or TCGA signature, in this resections,2 the authors suggested D2 dissection without and other studies, could provide the opportunity to splenopancreatectomy for these patients. Findings from individualise treatment in future studies. In study a 2017 randomised controlled trial26 supported the non- populations from Europe and North America, many inferiority of spleen preservation versus splenectomy in patients are not able to have all planned postoperative patients with proximal gastric cancer, in terms of overall treatment, and both tumour downstaging and survival. Such an approach (ie, D1+, D2, or D3 dissection) resectability are related to overall survival. Therefore, in was done in more than 544 (85%) of our patients. the successor trial (CRITICS-II; NCT02931890), all In our subgroup analysis in this study, the interaction treatment, consisting of docetaxel-based chemotherapy, between sex and treatment group was significant. The chemoradiotherapy, or both, will be administered before reason for this is unclear. In the Intergroup study,4 a surgery. univariate subgroup analysis showed a difference in Contributors survival according to sex, but the multivariate analysis AC, EPMJ, NCTvG, HB, CJHvdV, and MV contributed to the study could not verify this observation. concept. AC, EPMJ, NCTvG, HB, CJHvdV, MV, JWvS, HHH, HP, and HvT contributed to study design. AKT and HAMS contributed to study In this study, postoperative toxicity was similar in the coordination. AC, EPMJ, NCTvG, HB, CJHvdV, MV, KS, PL, MN, EM-KK, chemotherapy and chemoradiotherapy groups. Only AKT, HAMS, JWvS, and HHH contributed to data collection. AC, EPMJ, neutropenia (including grade 3–4 neutropenia) was more NCTvG, HB, CJHvdV, MV, KS, PL, MN, EM-KK, JWvS, HHH, HP, HvT, frequent in the chemotherapy group, but this did not HWMvL, and MIvBH contributed to data interpretation. All authors wrote the manuscript. AC, EPMJ, NCTvG, HB, CJHvdV, and MV edited translate into more febrile neutropenia and was therefore the manuscript. deemed clinically irrelevant by patients’ treating physicians. Declaration of interests A limitation of the CRITICS study is that upfront AC and MV received institutional financial support as investigators of randomisation does not allow a direct comparison of this study from the Dutch Cancer Society, the Dutch Colorectal Cancer postoperative chemotherapy with chemoradiotherapy. Group, and Hoffman-La Roche. AC has received support as an Because only 60% of all randomised patients would be investigator for study conduct from Nordic Pharma and has been a consultant for Eli Lilly and Merck. MIvBH declares a consultant or included in such an analysis, all properties of advisory role for Medtronic and Olympus, and research funding from randomisation are lost, precluding an unbiased Olympus. All other authors declare no competing interests. investigation of a causal effect of treatment. Another Acknowledgments limitation of our study is the low percentage of Funding from the Dutch Cancer Society (CKTO 2006-02; NKI 2006-4167) diagnostic laparoscopies (10%) done before was used to support central and local data management. The Dutch randomisation. In the Netherlands the first national Colorectal Cancer Group funded the on-site monitoring. An educational 27 grant from Hoffmann-La Roche was used to support site initiation visits guideline for gastric cancer, in 2009, advocated a and study-related costs for administration and transport of study diagnostic laparoscopy only in case of suspicion of material. We thank all patients, treating physicians, and research support peritoneal disease or poorly differentiated cT3–4 gastric staff in all participating centres; the Netherlands Cancer Institute Trial tumours. The current guideline of 201627 demands a Office for Safety Desk support and Monitoring; and the Central Datacenter of the Department of Surgery at the Leiden University diagnostic laparoscopy for all cT3–4 tumours according Medical Center, Netherlands, for central data management. to endoscopy and CT scan. Although the low diagnostic References laparoscopy rate might implicitly have affected the 1 Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A. dropout rate after surgery, the dropout rate in our study Global cancer statistics, 2012. CA Cancer J Clin 2015; 65: 87–108. does not exceed that of other studies investigating 2 Songun I, Putter H, Kranenbarg EM, Sasako M, van de Velde CJ. Surgical treatment of gastric cancer: 15-year follow-up results of the perioperative chemotherapy, including those with a randomised nationwide Dutch D1D2 trial. Lancet Oncol 2010; mandatory diagnostic laparoscopy (36% in the 11: 439–49. Intergroup study and 58% in the MAGIC study).6,16,18,19 3 Sasako M, Sano T, Yamamoto S, et al. D2 lymphadenectomy alone or with para-aortic nodal dissection for gastric cancer. In conclusion, we did not find better efficacy of N Engl J Med 2008; 359: 453–62. postoperative chemoradiotherapy compared with 4 Macdonald JS, Smalley SR, Benedetti J, et al. Chemoradiotherapy postoperative chemotherapy in patients with resectable after surgery compared with surgery alone for adenocarcinoma of the stomach or gastroesophageal junction. N Engl J Med 2001; gastric cancer treated with preoperative chemotherapy 345: 725–30. and adequate surgery. Tolerability was also similar 5 Smalley SR, Benedetti JK, Haller DG, et al. Updated analysis of between the two adjuvant regimens. Future explorative SWOG-directed intergroup study 0116: a phase III trial of adjuvant radiochemotherapy versus observation after curative gastric cancer post-hoc analyses might identify subgroups that benefit resection. J Clin Oncol 2012; 30: 2327–33. more from one of these strategies, although the findings 6 Cunningham D, Allum WH, Stenning SP, et al. Perioperative of such exploratory analyses should be interpreted with chemotherapy versus surgery alone for resectable gastroesophageal caution. Docetaxel might be a better chemotherapy cancer. N Engl J Med 2006; 355: 11–20. 17,19 7 Cunningham D, Starling N, Rao S, et al. Capecitabine and backbone than epirubicin and might therefore be a oxaliplatin for advanced esophagogastric cancer. N Engl J Med 2008; better option in future trials. Furthermore, classification 358: 36–46. of gastric cancer on the basis of histological response, 8 Dikken JL, van Sandick JW, Swellengrebel HA, et al. Neo-adjuvant chemotherapy followed by surgery and chemotherapy or by surgery completeness of surgical resection, or nodal positivity, and chemoradiotherapy for patients with resectable gastric cancer as well as molecular subtypes, such as HER2 (CRITICS). BMC Cancer 2011; 11: 329. www.thelancet.com/oncology Vol 19 May 2018 627 Articles

9 Greene FL PD, Fleming ID, Fritz A, Balch CM, Haller DG, 19 Al-Batran S-E, Homann N, Schmalenberg H, et al. Perioperative Morrow M, eds. American Joint Committee on Cancer: AJCC chemotherapy with docetaxel, oxaliplatin, and fluorouracil/ cancer staging manual, 6th edn. New York, NY: Springer, 2002. leucovorin (FLOT) versus epirubicin, cisplatin, and fluorouracil or 10 Pocock SJ, Simon R. Sequential treatment assignment with capecitabine (ECF/ECX) for resectable gastric or gastroesophageal balancing for prognostic factors in the controlled clinical trial. junction (GEJ) adenocarcinoma (FLOT4-AIO): a multicenter, Biometrics 1975; 31: 103–15. randomized phase 3 trial. Proc Am Soc Clin Oncol 2017; 11 Bonenkamp JJ, Hermans J, Sasako M, van De Velde CJ. 35 (suppl 15): 4004 (abstr). Quality control of lymph node dissection in the Dutch randomized 20 Alderson D, Cunningham D, Nankivell M, et al. Neoadjuvant trial of D1 and D2 lymph node dissection for gastric cancer. cisplatin and fluorouracil versus epirubicin, cisplatin, and Gastric Cancer 1998; 1: 152–59. capecitabine followed by resection in patients with oesophageal 12 Jansen EP, Boot H, Dubbelman R, Bartelink H, Cats A, Verheij M. adenocarcinoma (UK MRC OE05): an open-label, randomised Postoperative chemoradiotherapy in gastric cancer—a phase I/II phase 3 trial. Lancet Oncol 2017; 18: 1249–60. dose-finding study of radiotherapy with dose escalation of cisplatin 21 Fuchs CS, Niedzwiecki D, Mamon HJ, et al. Adjuvant and capecitabine chemotherapy. Br J Cancer 2007; 97: 712–16. chemoradiotherapy with epirubicin, cisplatin, and fluorouracil 13 Jansen EP, Boot H, Dubbelman R, Verheij M, Cats A. Postoperative compared with adjuvant chemoradiotherapy with fluorouracil and chemoradiotherapy in gastric cancer-a phase I-II study of leucovorin after curative resection of gastric cancer: results from radiotherapy with dose escalation of weekly cisplatin and daily CALGB 80101 (Alliance). J Clin Oncol 2017; 35: 3671–77. capecitabine chemotherapy. Ann Oncol 2010; 21: 530–34. 22 Kim S, Lim do H, Lee J, et al. An observational study suggesting 14 Jansen EP, Boot H, Saunders MP, et al. A phase I–II study of clinical benefit for adjuvant postoperative chemoradiation in a postoperative capecitabine-based chemoradiotherapy in gastric population of over 500 cases after gastric resection with D2 nodal cancer. Int J Radiat Oncol Biol Phys 2007; 69: 1424–28. dissection for adenocarcinoma of the stomach. Int J Radiat Oncol Biol Phys 2005; 63: 1279–85. 15 Grambsch P, Therneau T. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994; 81: 515–26. 23 Lee J, Lim do H, Kim S, et al. Phase III trial comparing capecitabine plus cisplatin versus capecitabine plus cisplatin with concurrent 16 Ychou M, Boige V, Pignon JP, et al. Perioperative chemotherapy capecitabine radiotherapy in completely resected gastric cancer with compared with surgery alone for resectable gastroesophageal D2 lymph node dissection: the ARTIST trial. J Clin Oncol 2012; adenocarcinoma: an FNCLCC and FFCD multicenter phase III trial. 30: 268–73. J Clin Oncol 2011; 29: 1715–21. 24 Claassen YHM, de Steur WO, Hartgrink HH, et al. 17 Al-Batran SE, Hofheinz RD, Pauligk C, et al. Histopathological Surgicopathological quality control and protocol adherence to regression after neoadjuvant docetaxel, oxaliplatin, fluorouracil, and lymphadenectomy in the CRITICS gastric cancer trial. Ann Surg 2017; leucovorin versus epirubicin, cisplatin, and fluorouracil or published online Aug 16. DOI:10.1097/SLA.0000000000002444. capecitabine in patients with resectable gastric or gastro-oesophageal junction adenocarcinoma (FLOT4-AIO): 25 Mocellin S, McCulloch P, Kazi H, Gama-Rodrigues JJ, Yuan Y, results from the phase 2 part of a multicentre, open-label, Nitti D. Extent of lymph node dissection for adenocarcinoma of the randomised phase 2/3 trial. Lancet Oncol 2016; 17: 1697–708. stomach. Cochrane Database Syst Rev 2015; 8: Cd001964. 18 Cunningham D, Stenning SP, Smyth EC, et al. Peri-operative 26 Sano T, Sasako M, Mizusawa J, et al. randomized controlled trial to chemotherapy with or without bevacizumab in operable evaluate splenectomy in total gastrectomy for proximal gastric oesophagogastric adenocarcinoma (UK Medical Research Council carcinoma. Ann Surg 2017; 265: 277–83. ST03): primary analysis results of a multicentre, open-label, 27 Oncoline. Landelijke Richtlijn maagcarcinoom 2.2. 2017. randomised phase 2–3 trial. Lancet Oncol 2017; 18: 357–70. http://www.oncoline.nl/maagcarcinoom (accessed March 29, 2018).

628 www.thelancet.com/oncology Vol 19 May 2018